I'm not sure favorability is junk science so much as it is not a good predictor of who will win delegates through a series of byzantine and varied state processes.

And at the risk of stating the obvious, finding someone as more favorable than their opponent doesn't necessarily imply I will vote for them in an election.